Making Games with Agents

You have a game in your head. You've had it for years. You think about it on your commute, in the shower, while you're supposed to be doing something else. You know what it feels like to play. You can hear the music. You've sketched the characters on napkins and in notebooks and in the margins of meeting agendas.

But you haven't built it. Because building a game takes a team, and a team takes funding, and funding takes a prototype, and a prototype takes weeks of setup before you write a single line of gameplay code. So the game stays in your head.

I've built dozens of prototypes. Self-published games. Then helped publish indie games at Devolver Digital. Tons of pitches, just as many prototypes. The ones that got most of us excited always came down to the same thing: a passionate team, a singular vision, and a memorable build you could put in someone's hands. Not a deck. Not a pitch doc. Something you could feel.

The problem was never the idea. The problem was always the distance between the idea and something real.

That distance just got a lot shorter.

The funding reality

If you're making games in Canada, you already know about the Canada Media Fund.^[1] You know the programs are competitive, the process can be opaque, and the timelines don't match how games actually get made. You apply, you wait, you might get a fraction of what you asked for. Meanwhile the clock is ticking on your savings, your day job patience, your partner's goodwill.

Even when the money comes through, the math is brutal. A small team burning runway just to get a vertical slice. A good chunk of that time is setup--choice of engine, build pipelines, asset workflows, networking stubs, UI scaffolding. The stuff that has to exist before you can test the one mechanic that makes your game worth playing.

What if that setup phase shrank dramatically? Not because the work disappeared, but because an agent handled it while you focused on the part only you can do--the design, the feel, the thing that makes your game yours.

The typical team

Most indie teams start with a mix of the following people:

• Developer--writes the code. Gameplay, systems, tools, builds. Everything that runs.
• Artist--creates the assets. Characters, environments, UI, animation. Everything you see.
• Designer--designs the experience. Mechanics, levels, systems, balance. Everything you play.
• Musician--handles the audio. Music, sound design, adaptive scores, mixing. Everything you hear.
• Writer--builds the world. Narrative, dialogue, lore, item descriptions. Everything you read.
• Marketer--runs the presence. Social media, Discord, devlogs. Everything that gets you noticed.
• Producer--keeps the project moving. Schedules, budgets, submissions. Everything behind the scenes.

Separate lanes, separate worlds.^[2] The developer doesn't touch the art pipeline. The artist doesn't write shaders. The writer doesn't know how dialogue gets wired into the engine. And so on. Everyone stays where they're comfortable, and every dependency between disciplines becomes a bottleneck. Every bottleneck becomes a delay. Every delay chips away at the runway.

Reaching across

AI tools give your team the reach to work across disciplines. Not by replacing expertise, but by lowering the barrier to everything adjacent to it.

• Developer--now generates concept art for prototyping, creates placeholder music and sound effects, drafts store copy. Still delivers all the code--but no longer blocked waiting for assets.
• Artist--now writes tool scripts, builds asset pipelines, automates export workflows. Still creates all the art--but no longer dependent on the developer for integration.
• Designer--now prototypes mechanics directly, builds level editors, tests balance changes live. Still designs the game--but can validate ideas without waiting on a build.
• Musician--now builds adaptive audio systems, prototypes sound-reactive visuals, generates scratch tracks for trailers. Still owns the soundtrack--but can ship audio features without waiting on engineering.
• Writer--now builds branching dialogue tools, scripts cutscene prototypes, generates localization drafts. Still writes the story--but can test narrative flow end-to-end without a developer.
• Marketer--now generates trailer cuts, builds Discord bots, ships analytics dashboards. Still runs the community--but can move at the speed of the conversation.
• Producer--now vibes prototypes, automates platform submissions, generates status reports from commit history. Still keeps everyone on track--but with technical leverage that used to require a developer.

Every specialist becomes a generalist who's still a specialist. That's the compound effect--you don't add capability, you multiply it.^[3]

The elephant in the room

I know what you're thinking. Some of you have been thinking it since the title. “AI tools destroy jobs and ruin livelihoods.”

I understand why people say that. But I'm not talking about what corporations will do with this technology. I'm talking about a small team with no funding and no safety net, trying to build something on weekends. People who can't afford to hire a second developer. People who can't afford a $15,000 bootcamp to learn a new discipline. People whose game will never exist without leverage they don't currently have.

There's a learning curve. These tools reward structured thinking, and not everyone on your team will take to them at the same pace. But the cost of learning is time, not money--and the tools get better at meeting you where you are.

As of today, the most capable AI agents cost between $20 and $200 a month--no more than $2,400 a year. Less than a week of contractor time. Less than what you pay to attend GDC or PAX or BitSummit. And they teach you on the job--for example, every time an agent writes code, you can read it, ask questions, understand the approach. It's a tutor, a pair programmer, and a force multiplier all at once.

The game in your head is closer than you think. The tools exist. The cost is negligible. And when the funding conversation happens, you show up with a build, not a deck. The only thing between your team and a playable prototype is the decision to start.

Go make something. Grab your team, enter a game jam, build the worst version of your best idea. Let an agent handle the scaffolding while you focus on the part that matters--the part that only your team can make.